On 08/23, Andrii Nakryiko wrote: > > This is a bit confusing, because even if the kernel-side uretprobe > handler doesn't do the filtering by itself, uprobe subsystem shouldn't > install breakpoints on processes which don't have uretprobe requested > for (unless I'm missing something, of course).
Yes, but in this case uprobe_register/apply will install breakpoints on both tasks, with PID1 and PID2. > It still needs to be fixed Agreed, > like you do in your patch, Not sure... > more, we probably need a similar UPROBE_HANDLER_REMOVE handling in > handle_uretprobe_chain() to clean up breakpoint for processes which > don't have uretprobe attached anymore (but I think that's a separate > follow up). Probably yes... but yes, this is another issue. Oleg.
